Daemon or no daemon

Magnus Bergman magnus.bergman at observer.net
Mon Apr 18 20:31:53 EEST 2005


On Mon, 18 Apr 2005 12:49:16 -0400
Sean Middleditch <elanthis at awesomeplay.com> wrote:

> > don't understand: "The complexity in making backends support both
> > in-process and daemon operation would severely complicate the
> > development of a backend." Why should that be so complicated? If it
> 
> Because backends in the daemon will be dependent on being in their own
> threads, with a management thread, and threads per connecting
> application, and probably even some more threads, all because of the
> wonderfully stupid synchronous APIs we have to build on top of.
> Actually putting those in the client will require the entire daemon,
for
> the most part, to be reimplemented in the client library to manage all
> that gook, all for next to no actual real benefit.

I was not aware of any of the threading issues (as I hoped some kind of
main-loop solution instead). But I was aware of the massive code
duplication issue, which was exactly what I addressed with my design
idea: Moving that code to the (client) library and have the daemon use
that same library, then there will be no code duplication. You are
probably right that there will be no big benefit (in the D-VFS case) and
that has to be balanced against the extra work needed to be done of
course (and I might be too optimistic about that).



More information about the xdg mailing list